After setting up Microsoft's Surface Laptop Ultra, the first reaction is not wonder at the aluminum chassis or the bright display. It is a far more practical question: what would anyone actually do with a $5,900 laptop?
This is not an ordinary Windows machine. Microsoft partnered with Nvidia to build a high-end notebook designed to showcase Nvidia's new Arm-based RTX Spark chips. It marks a significant moment for Windows on Arm, because these processors can run demanding AI models locally while also handling the latest AAA games and creative applications. The pitch is compelling. The reality of local AI is considerably messier.
Getting the most out of the Surface Laptop Ultra, particularly the top-of-the-line configuration, requires coding experience or at least the patience to experiment with models and AI agents until something works. For that specific audience, Microsoft may have built the Windows laptop they have been waiting for. For everyone else, it is a much harder sell.
A design that finally rivals Apple
It feels as though Microsoft has finally built hardware capable of competing with Apple's professional laptops. The all-aluminum design, large haptic trackpad, and high-end finish give the machine a premium feel. Even the nightfall color closely echoes Apple's space black finish.
There are meaningful differences, however. The 15-inch mini-LED display can reach up to 2,000 nits of peak HDR brightness at 262 pixels per inch, with a resolution of 3270 x 2180. That makes it the brightest display Microsoft has ever shipped on a Surface Laptop. Unlike the MacBook Pro, it is also a 3:2 touchscreen, which suits writing code, reading text, and editing photos particularly well.
The port selection is broadly similar to the MacBook Pro, with one very notable change. There are three USB-C ports, a USB-A port, a single HDMI 2.1b port, an SD card reader, and a headphone jack. Where it truly differs is Microsoft's new Magnetic Connect. It works much like the old Surface Connect, but through the USB-C port on the right-hand side. Bring the connector close to the port and it snaps into place. When not charging, that port behaves like any other USB-C port on the machine.
Magnetic Connect is one of the best additions on the Surface Laptop Ultra. The magnetic USB-C charging end can be reversed and attached to the power brick, allowing the regular USB-C connector at the opposite end of the cable to charge other devices. It is a small design decision that adds real everyday convenience.
The haptic feedback on the large trackpad is another welcome addition. It responds to clicks, drags, and window alignment throughout Windows 11, and the tactile feedback also triggers in applications such as PowerPoint when dragging and aligning objects or images in a slide deck. The effect is subtle enough to make it feel as though windows and interface elements are genuinely being moved, though wider app support would be welcome.
Microsoft has included a 1080p front-facing camera above the screen. Image quality is fine for video calls, and it supports Windows Hello face authentication along with Windows Studio Effects such as automatic framing, eye contact correction, and background blur.
Rough edges in everyday use
Not all of the hardware impresses. There have been instances where unplugging the charger at 100 percent caused Windows to immediately report 97 percent battery life. Plugging the charger back in would return the indicator to 100 percent, only for it to drop to 97 percent again the next time the cable was removed.
According to Microsoft, this behavior is intentional. When the device is plugged in and the battery is full, charging pauses rather than constantly topping off, and it resumes only once the level drops to 95 percent. The goal is to reduce battery wear over time. While plugged in, the indicator shows 100 percent to signal a full charge; when unplugged, it shows the exact level right away, which may be a few percentage points lower.
The display has also produced some odd behavior. At times, interactions felt slightly laggy, and there was noticeable ghosting. Rapidly moving a window around could leave a small trail. Switching power settings from the default best power efficiency mode to best performance appeared to resolve most of the laggy feeling, and the backlight ghosting became less noticeable the longer the laptop ran.
The built-in speakers have also produced loud pops and bangs. They were loud enough to sound like firecrackers, briefly raising concern that the battery was about to fail before it became clear the noise was speaker interference. Rebooting solved the problem, but it has occurred more than once.
Rebooting itself is surprisingly slow, which seems like something Microsoft could address with a software update.
Local AI models and agents
The RTX Spark chip at the heart of the Surface Laptop Ultra is designed with local AI models and agents in mind. The tested variant has a 20-core CPU codesigned by Nvidia and MediaTek paired with a 6,144-CUDA-core GPU, roughly comparable to a mobile RTX 5060, plus 128GB of unified memory and a 1TB Gen 5 SSD. That top configuration costs $5,899. The base version starts at $2,599 and comes with an 18-core CPU, a 5,120-CUDA-core GPU, and 24GB of unified memory.
Out of the box, Microsoft prompts owners to pick an AI agent, with options including Hermes, OpenClaw, and Perplexity. Getting OpenClaw to work initially required a reinstall to fix a gateway issue, after which an agent could be named and used for local AI tasks.
Both Hermes and OpenClaw were configured to use the Qwen3.8-27B model, one of the more powerful local AI models available. Because it draws on the full power of the machine, the fans spin up at full throttle and the GPU runs at 100 percent even for basic queries. Asking Hermes to resize and convert some photos from PNG to JPG, a task that takes seconds with a simple utility, took minutes to complete while the fans buzzed and the laptop grew hot.
Part of that is down to choosing such a large model for a basic task, but without testing every local model available, it is difficult to know which one suits each job. OpenClaw was asked to build a smart home dashboard and spent 30 minutes spinning up the fans and calling various tools before producing nothing but an error.
The base model's 24GB of unified memory will clearly limit how many local AI models users can choose from, even without hands-on testing.
Compelling use cases for this level of local AI power were hard to find. Many tasks can be completed far more quickly, and without the fans spinning up, by using a free cloud model. The real advantage lies in leaving an agent running at full throttle to build something overnight or to process large datasets that would otherwise cost a lot in cloud tokens.
Local processing does offer privacy benefits. Feeding Hermes audio and video files led it to create a local llama.cpp server for audio transcription. It took some fiddling to get working, but it could save the cost of installing transcription apps or paying for a cloud service.
The main drawback is that these agents are far from user-friendly. It is surprising that neither Microsoft nor Nvidia has developed a local AI interface that simplifies the process. As someone who does not code, it is clear this is all targeted squarely at developers and power users.
Gaming and creative tasks
The Surface Laptop Ultra may not be an official gaming laptop, but it can run far more of the latest games than Apple's MacBook Pro. Windows on Arm gaming has traditionally been a concern given the weak GPU performance of Qualcomm's Snapdragon X series chips. This is the first time an Arm CPU has been paired with a stronger Nvidia GPU, and most games run well for a laptop that is not the size of a gaming machine.
Because this is an RTX GPU, it supports DLSS, Frame Generation, and ray tracing in full. Frame Generation can be enabled to reach 120fps and match the 120Hz display. Black Myth: Wukong ran at an average of 70fps at native resolution with medium settings and DLSS performance mode. Cyberpunk 2077 averaged 66fps on the same settings, while Forza Horizon 6 managed 57fps.
Most games need medium settings and DLSS performance or balanced mode to reach around 60fps, though there are exceptions. Gears of War: E-Day averaged 40fps on medium settings and DLSS performance, rising only to 48fps on low settings. Doom: The Dark Ages required low settings and DLSS performance to hit an average of 60fps.
Game compatibility is good for most titles, but emulation issues appear in some. Horizon Zero Dawn Remastered was limited to 30fps at 1440p regardless of whether low or medium settings were used, likely because the game is CPU-heavy and would benefit from a native Arm port.
Native Arm games remain rare. The Arm version of Marvel Rivals averaged 64fps with medium settings and DLSS performance, and 75fps on low settings. More Arm ports are likely for games with anti-cheat engines, which are far more complicated to emulate. Titles such as Arc Raiders, Counter-Strike 2, Dota 2, and Fortnite work out of the box, while Nvidia and Microsoft are working with EA, Activision, and Riot Games to support Battlefield, Call of Duty, Valorant, and League of Legends.
For creative work, the machine is capable. A 4K video test exported in two minutes and 18 seconds using Premiere Pro, around a minute behind a top-of-the-line MacBook Pro but ahead of a base M5 chip and roughly 30 seconds behind an M5 Pro. The export had to be run through Premiere Pro's own tool rather than Media Encoder because of a bug in the beta Arm version of Media Encoder.
Where the Surface Laptop Ultra falls furthest behind is CPU performance. Single-core and multi-threaded tests such as Geekbench and Cinebench show how much weaker the MediaTek CPU is against an M5 Max or even an M6 Mac Mini. Some GPU tests also show the RTX Spark trailing Apple's hardware, and Photoshop, Premiere Pro, and DaVinci Resolve benchmarks are all noticeably behind the M5 Max.
Battery life appears to be around seven hours when using local AI models alongside a mix of apps and gaming. With less intensive work, it averages 10 hours or more.
Outside of some AI workloads such as image generation and native CUDA compute, the Surface Laptop Ultra does not offer a clear performance advantage over the MacBook Pro. It is not transformative for the Windows world in the way the M1 chips were for macOS, largely because Qualcomm already pushed Windows on Arm forward with its Snapdragon X series two years earlier. That makes the machine difficult to recommend unless the buyer is specifically interested in running local AI models, coding, or experimenting with AI agents.
The Surface Laptop Ultra and Nvidia's RTX Spark platform target developers and power users who want a laptop that can handle those workloads alongside creative apps and the latest games. For that audience, there is a great deal to like. For everyone else, more affordable laptops can handle everyday productivity, creative work, and some light gaming without the hefty price premium. Microsoft has come closer than ever to building a Windows alternative to the MacBook Pro, with hardware that looks and feels excellent and an Nvidia GPU that brings CUDA support and decent gaming to a high-end design. But the machine's biggest selling point is also its biggest question mark: local AI still feels like a work in progress, and the price of experimenting is steep.
Source:The Verge News